critic-review
Skill6 starsChanged 20 days agoMITView on GitHub
Review an engineering design, code change or result against requirements and raw evidence. Use for independent correctness, security, reliability, performance and evidence review before integration or consequential claims.
2 other projects use a version of this.
Adopted by
Most recently taken up by
- samuhs/mad-sciencetheir own version, first seen Nov 1, 2026
- junhjang/claude-harnesstheir own version, first seen May 12, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- samuhs/mad-science★ 1same name
- junhjang/claude-harness★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository10 other files it gives its agents.
- AGENTS.mdAGENTS.md
- claim-ledgerSkill
- distributed-orchestrationSkill
- engineering-governanceSkill
- engineering-planSkill
- execution-integritySkill
- literature-provenanceSkill
- multi-agent-collaborationSkill
- technical-writingSkill
- worker-taskSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 6
Read requirements, patch and relevant outputs yourself.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Read requirements, patch and relevant outputs yourself. Review the actual revision being integrated. Another role's summary is not proof; do not edit unrelated code to manufacture a finding.
Under Milestone Mode, batch reviews at Pilot Readout, Pre-Main Launch and CLAIMS Entry using one frozen evidence packet. Do not join routine computation or schedule hourly model reviews. H is the final project decision authority within host constraints: document a material objection once, then respect a trusted H decision. Distinguish an accepted risk/waiver from a passed test and never upgrade evidence merely because H chose to proceed. An unresolved scope/resource disagreement blocks to H through O. C reviews implementation; blind V must receive its minimal packet without that implementation.
Apply relevant lenses:
- Acceptance: does each required condition pass, including boundaries and failures?
- Correctness: state transitions, concurrency, invariants, compatibility and data integrity.
- Failure handling: swallowed exceptions, fallbacks, partial success and misleading completion.
- Security: trust boundaries, authorization, credentials and injection in changed paths.
- Performance: measured workload, resource limits, comparable baselines and realistic cost.
- Evidence: selection bias, repeated comparisons, skipped failures, uncertainty and overclaiming.
Estimate confidence that each candidate is a real, material defect. Report findings at 80–100 with severity, precise location, failing condition and remedy. Keep 50–79 as labeled notes, and lower-confidence points only as specific cheap checks. Confidence is separate from severity; a speculative severe scenario is not a confirmed finding.
State whether acceptance criteria are met, corrections are needed or the result remains unverified. "No actionable findings" is valid. Give the owner a reproducer rather than silently changing files outside your assignment. Review does not authorize merge, publication or deployment.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

